Amazon Web Services announced a series of updates to its Recognition service, which provides machine learning-based computer vision capabilities to cloud customers.
The system will now be able to detect and recognize text in images, so customers can feed in signs and documents and get the contents of those images back for processing further. That means Recognition can be used for making images of the physical world more intelligible by systems that are only built for processing textual data.
Customers will also be able to perform real-time face searches across collections of millions of faces. For example, Rekognition could be used to verify that an image of a person matches another one on file in an existing database of up to tens of millions of images, with sub-second latency.
That’s particularly useful in law enforcement cases, when customers want to match one person’s photo with existing images already on file. (It also raises questions about the ethics of a technology company building tools that law enforcement can use to pursue criminals.)
On top of all that, the system can also detect up to 100 different faces in a photo, compared to its previous limit of 15 faces. That means companies can feed Rekognition a shot of a crowd of people, and get information back about the demographics of all the faces detected, as well as what emotions the system believes they’re displaying.
The changes are part of an ongoing war between cloud providers including AWS, Google, Microsoft, IBM, and Salesforce. All of these businesses are trying to make the machine learning revolution more accessible by providing customers with pre-built systems that promise intelligent results.
In addition, systems like Rekognition can help attract customers to different cloud platforms, bringing along more workloads and keeping businesses engaged for long periods of time.
This news comes a week before AWS re:Invent, the company’s massive conference for customers and partners in Las Vegas, Nevada. AWS is expected to launch a whole suite of new and updated products and services at that event, so stay tuned.
